B1446
Title: R-vines as a new way to model interactions within French dairy-cattle systems
Authors: Naomi Ouachene - Institut Agro - INRAE (France) [presenting]
Claudia Czado - Technical University of Munich (Germany)
Tristan Senga Kiesse - Institut Agro - INRAE (France)
Michael Corson - Institut Agro - INRAE (France)
Abstract: In the context of climate change, increasing the environmental performances of farms without compromising productivity guarantee of food security and farm revenue is a major issue. A farm emits several types of greenhouse gases, whose multiple sources are connected by complex dependence structures, which make it hard to model. This raises the issue of how to adequately represent the multiple interactions of farm descriptive variables to contribute to a better understanding of systems and improve their performances. To address this issue, regular vine copulas are investigated for their ability to map multivariate complex dependencies by taking advantage of the large variety of bivariate copulas as building blocks of their tree structure. The method was applied to a dataset which describes management practices, emissions and productivity of French dairy farms. The approach offered a new way to represent farms as a function of a set of variables. A first analysis, including all the farms, identified the specificities of different kinds of systems. A second assessment, per type of system, allowed for a deeper understanding of the impact of different practices according to the farm context and their role in the improvement of the performances of farms.
